Re: [PATCH] stdlib/longlong.h: Remove lvalue to rvalue conversion

[Florian Weimer <fw@deneb.enyo.de>]

* Fāng-ruì Sòng:

> On Mon, Oct 11, 2021 at 11:37 PM Florian Weimer <fw@deneb.enyo.de> wrote:
>>
>> * Fangrui Song via Libc-alpha:
>>
>> > An output constraint takes a lvalue. While GCC happily strips the
>> > incorrect lvalue to rvalue conversion, Clang Clang rejects the code by
>> > default:
>> >
>> >     error: invalid use of a cast in a inline asm context requiring an lvalue: remove the cast or build with -fheinous-gnu-extensions
>>
>> As I wrote on the gcc-patches list, these casts may be needed in a few
>> cases.  Just removing them may not be entirely correct.
>
> You meant https://gcc.gnu.org/pipermail/gcc-patches/2021-October/581265.html
>
> "This seems to alter the meanining of existing programs if sh and sl do
> not have the expected type.
> I think you need to add a compound expression and temporaries of type
> USItype if you want to avoid the cast."
>
>
> I tested this on x86-64 and aarch64 and saw no new failure.

It's probably not possible to show conclusively using testing alone
that the change is safe.  A type mismatch can only show up with
certain GCC versions or optimization levels.

The definitions of mp_limb_t and USItype are unfortunately different,
so it is hard to tell if all uses of these macros are still correct.

> In addition, using a compound statement sorta defeats the purpose to
> simplify the code.

I thought the actual goal was to get this to compile with Clang?
Clang does not support typed asm output operands, so the code is going
to turn more complex.